TL;DR本文研究了视觉检测任务的增量学习,通过一种在线流式学习的方法,使用一种新颖的记忆重放机制,使系统可以在时间上逐步引入新类别来完成物体检测任务,并在 Pascal VOC 2007 和 MS COCO 数据集上取得了最好的实验结果。
Abstract
Humans can incrementally learn to do new visual detection tasks, which is a
huge challenge for today's computer vision systems. Incrementally trained deep
learning models lack backwards transfer to previously seen classes and suffer
from a phenomenon known as $"catastrophic forgetting.